RedPajama-INCITE-Base-3B-v1: benchmark results
Together's RedPajama-INCITE-Base-3B-v1, a 2.8B pretrained language model developed with Ontocord.ai, ETH DS3Lab, Stanford CRFM and LAION, trained on 3,072 V100 GPUs under the INCITE 2023 project. Provider: Together. Released 2023-05-04. Access: Open.
Aggregate Rating 1152 ± 13, rank #1600 of 1640 rated models, from 102 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - High School Physics | 37.09 | Accuracy (%) | 88.1 |
| MMLU-by-task - College Mathematics | 36 | Accuracy (%) | 80.2 |
| HELM Classic - Entity Data Imputation | 82.82 | Exact Match (%) | 78.8 |
| HELM Classic - TruthfulQA | 27.68 | Exact Match (%) | 68.2 |
| HELM Classic - LSAT | 21.3 | Exact Match (%) | 66.9 |
| MMLU-by-task - Moral Scenarios | 26.59 | Accuracy (%) | 55.1 |
| HELM Classic - CivilComments | 54.87 | Exact Match (%) | 54.5 |
| MMLU-by-task - Abstract Algebra | 29 | Accuracy (%) | 45.4 |
| MMLU-by-task - High School Statistics | 36.57 | Accuracy (%) | 45.4 |
| HELM Classic - Dyck | 53 | Exact Match (%) | 44.1 |
| MMLU-by-task - Elementary Mathematics | 26.98 | Accuracy (%) | 39.2 |
| HELM Classic - bAbI | 46.58 | Exact Match (%) | 39.1 |
Interactive version: theaggregate.ai/model?slug=redpajama-incite-base-3b-v1 · How It Works · Data refreshed daily, snapshot 2026-11-07.